ieee access 1 crowdsensim: a simulation platform for

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. Seehttp://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2671678, IEEE Access

IEEE ACCESS 1

CrowdSenSim: a Simulation Platform for MobileCrowdsensing in Realistic Urban Environments

Claudio Fiandrino, Student Member, IEEE, Andrea Capponi, Student Member, IEEE, Giuseppe Cacciatore,Dzmitry Kliazovich, Senior Member, IEEE, Ulrich Sorger, Pascal Bouvry, Member, IEEE,

Burak Kantarci, Senior Member, IEEE, Fabrizio Granelli, Senior Member, IEEE,and Stefano Giordano, Senior Member, IEEE

Abstract—Smart cities take advantage of recent ICT devel-opments to provide added value to existing public servicesand improve quality of life for the citizens. The Internet ofThings (IoT) paradigm makes the Internet more pervasivewhere objects equipped with computing, storage and sensingcapabilities are interconnected with communication technologies.Because of the widespread diffusion of IoT devices, applyingthe IoT paradigm to smart cities is an excellent solution tobuild sustainable Information and Communication Technology(ICT) platforms. Having citizens involved in the process throughmobile crowdsensing (MCS) techniques augments capabilitiesof these ICT platforms without additional costs. For properoperation, MCS systems require the contribution from a largenumber of participants. Simulations are therefore a candidatetool to assess the performance of MCS systems. In this paper,we illustrate the design of CrowdSenSim, a simulator for mobilecrowdsensing. CrowdSenSim is designed specifically for realisticurban environments and smart cities services. We demonstratethe effectiveness of CrowdSenSim for the most popular MCSsensing paradigms (participatory and opportunistic) and wepresent its applicability using a smart public street lightingscenario.

Index Terms—Mobile crowdsensing, simulations, smart cities.

I. INTRODUCTION

WORLD population living in cities has experienced anunprecedented growth over the past century. While

only 10% of the population lived in cities during 1900, todaythis percentage corresponds to 50% and it is projected to fur-ther increase beyond such figure [1]. Sustainable developmentplays therefore a crucial role in city development. While only2% of the world’s surface is occupied by urban environments,cities contribute to 80% of global gas emission, 75% of globalenergy consumption [2] and 60% of residential water use [1].

C. Fiandrino is with Imdea Networks Institute, Leganés, Madrid, Spain.E-mail: {claudio.fiandrino}@imdea.org. Claudio developed this work as aPh.D. student at the University of Luxembourg.

A. Capponi, U. Sorger, and P. Bouvry are with the University ofLuxembourg, Luxembourg. E-mail: {firstname.lastname}@uni.lu.

D. Kliazovich is with ExaMotive, Luxembourg. E-mail:[email protected].

G. Cacciatore, and F. Granelli are with the University of Trento, Italy.E-mails: [email protected], [email protected].

B. Kantarci is with the School of Electrical Engineering andComputer Science, University of Ottawa, Ottawa, ON, Canada. E-mail:{burak.kantarci}@uOttawa.ca.

S. Giordano is with the University of Pisa, Pisa, Italy.E-mail: [email protected].

Smart cities rely on Information and Communication Tech-nology (ICT) solutions to improve citizens’ quality of life [3],[4]. The application of the Internet of Things (IoT) paradigmto urban scenarios is of special interest to support the smartcity vision [4]–[6]. Indeed, IoT is envisioned as a candidatebuilding block to develop sustainable ICT platforms. WithIoT, everyday life objects become uniquely identifiable and“smart”, i.e., they are equipped with computing, storage andsensing capabilities and can communicate one with each otherand with the users to enable pervasive and ubiquitous comput-ing [7]. Including citizens in the loop with crowdsensing ap-proaches augments the capabilities of existing infrastructureswithout introducing additional costs and has been proved tobe a win-win strategy for smart city applications [8]–[10].

Mobile crowdsensing (MCS) has emerged in the recentyears, becoming an appealing paradigm for sensing data [11].In MCS, users contribute data generated from sensors em-bedded in mobile devices, including smartphones, tablets andIoT devices like wearables. Accelerometer, gyroscope, magne-tometer, GPS, microphone and camera are just a representativeset of sensors which are nowadays employed to operate anumber of applications in many domains, including, among theothers, health care, environmental and traffic monitoring andmanagement [12], [13]. To illustrate with a simple example,Google exploits crowd-sourced information about smartphoneslocations to offer real-time view of congested traffic on roads,or its recently released Science Journal, which permits to col-lect and visualize data coming from smartphone sensors [14].

The information acquired through MCS platforms is usuallyaggregated and delivered to a collector typically located inthe cloud (see Fig. 1). This enables the so-called Sensingas a Service (S2aaS) model [5], which makes the collectedpublic data available to developers and end-users. With S2aaScompanies have no longer need to invest and acquire infras-tructure to perform a sensing campaign. IoT and MCS arekey enablers in the S2aaS model. Efficiency of S2aaS modelsis defined in terms of the revenues obtained from selling dataversus the costs of the sensing campaign, which include costsof recruitment and compensation of the participants for theirinvolvement [15]. Also, the users sustain costs while contribut-ing data. These costs correspond to the energy spent fromthe batteries for sensing and reporting data and, eventually,the data subscription plan if cellular connectivity is used forreporting.



IEEE ACCESS 2

S2aaS Cloud Collector

Participants

Mobile and IoT Devices

Data

LTE

WiFi

Accelerometer Gyroscope Microphone Dual Camera Temperature

Fig. 1. Cloud-based MCS system

In MCS, data acquisition or collection, can be participatoryor opportunistic [12]. In opportunistic sensing systems, theuser involvement is minimal: sensing decisions are application-or device-driven. In participatory sensing systems, users areactively engaged in the sensing process. The users, also calledparticipants in the remainder of the paper, are recruited by acentral platform, which dispatches sensing tasks. Users canthen decide which request to accept and, after accepting, theyhave to accomplish specified sensing and data reporting tasks.On one side, opportunistic sensing lowers the burden of userparticipation as devices or applications are responsible to takesensing decisions. Conversely, participatory sensing systemsare tailored to crowdsensing architectures with a “centralplatform”, which facilitates system control operations like taskassignment, user incentives and rewarding to compensate theparticipants for their contribution.

In this paper we propose CrowdSenSim, a new tool forsimulating mobile crowdsensing activities in realistic urbanenvironments. CrowdSenSim is specifically designed to per-form analysis in large scale environments and supports bothparticipatory and opportunistic sensing paradigms. CrowdSen-Sim allows scientists and engineers to investigate performanceof the MCS systems, with a focus on data generation andparticipant recruitment. The simulation platform can visualizethe obtained results with unprecedented precision, overlayingthem on city maps. In addition to data collection performance,the information about energy spent by participants for bothsensing and reporting helps to perform fine-grained systemoptimization.

The contribution synopsis of this paper is as follows:• Proposal of CrowdSenSim, a simulation platform for

MCS systems deployed in realistic urban environmentsand presentation of its design features.

• Validation of CrowdSenSim’s performance for oppor-tunistic and participatory sensing systems.

• Application of CrowdSenSim in a public street lightingscenario, an essential service in current and future smartcities.

The paper is organized as follows. Section II illustrates theexisting tools for simulation of MCS activities. Section IIIpresents the design criteria of CrowdSenSim, highlighting itsobjectives and scenarios of applicability. Section IV details

CrowdSenSim’s architecture. Section V presents performanceevaluation and Section VI illustrates the use of CrowdSenSimfor smart lighting. Finally, Section VII concludes the work andoutlines directions for future work on the topic.

II. BACKGROUND ON CROWDSENSING SIMULATIONTOOLS

Currently, existing simulation tools for MCS aim either atcharacterizing and modeling communication aspects or defineusage of spatial environment [16]. The following paragraphsoverview the main properties of each tool.

Tanas et al. propose to exploit Network Simulator 3 (NS-3)for crowdsensing simulations [17]. The objective is to assessperformance of a crowdsensing network taking into accountthe mobility properties of the nodes together with the wirelessinterface in ad-hoc network mode. Furthermore, the authorspresent a case study about how participants could reportincidents in the public rail transport. NS-3 provides highlyaccurate estimations of network properties. However, havingdetailed information on communication properties comes atthe expense of scalability. First, it is extremely difficult toperform simulations with a number of users contributing datain the order of tens of thousands. Second, the granularity ofthe duration of NS-3 simulations is typically in the orderof minutes. It reflects the objective to capture insights intothe behavior of communication protocols such as TCP, whichbecomes too detailed as typical duration of a sensing campaignis in the order of hours or days.

In [18], Farkas and Lendák present a simulation environ-ment developed to investigate performance of crowdsensingapplications in an urban parking scenario. Although the ap-plication domain is only parking-based, the proposed solutioncan be applied to other crowdsensing scenarios. The simulationscenario considers drivers as type of users that travel from oneparking spot to another one. The users are the sensors thattrigger parking events.

Mehdi et al. propose CupCarbon [19], which is a discrete-event wireless sensor network (WSN) simulator for IoT andsmart cities. One of the major strengths is the possibility tomodel and simulate WSN on realistic urban environmentsthrough OpenStreetMap. To set up the simulation, the re-searchers are required to individually deploy on the mapthe various sensors and the nodes such as mobile users, gasand media sensors and base stations. Therefore, the approachis suitable for experiments with scenarios comprising up tohundreds of nodes.

III. CROWDSENSIM: DESIGN PRINCIPLES

This section presents CrowdSenSim in a nutshell, high-lighting the principles of the design, its objectives and thescenarios of applicability. Performing simulations in complexenvironments, such as modern cities, requires the simulationplatform to be scalable. In other words, it should not limitthe researcher in the choice of key parameters such as thesimulation period or the number of users.

Scalability: For proper operation, MCS systems require alarge number of contributors. Therefore, CrowdSenSim is



IEEE ACCESS 3

designed to take into account participants in the order of tensof thousands that move in a wide realistic urban environment.Each individual can potentially own several mobile and IoTdevices. The time dimension is also important. The duration ofa sensing campaign can range from hours to days and Crowd-SenSim addresses this challenge efficiently. For instance, let usconsider 10 000 users producing data with a duration of only30 minutes per day. Using commonly available sensors onthe market like an accelerometer working at 50 Hz frequency12 bits long samples, the total amount of generated data byeach user would be 1.35 GiB. Considering the prolongedduration of the user contribution and additional sensors wouldconsiderably augment this figure.

Realistic urban environment: CrowdSenSim relies on real-istic urban environments, which makes the simulator flexibleand easy to be adopted in any city. Furthermore, it allows toperform analysis that provide meaningful insights to munici-palities to understand the feasibility and the potential of publicservices employing MCS techniques. Simulations over a gridor a square area as abstraction levels lower the complexity,but do not allow taking into account important features suchas movements in real streets and physical obstacles such asbuildings. CrowdSenSim incorporates this feature allowingusers to include the layout of cities as input.

User mobility: Human mobility is defined as sequences ofspatiotemporal user movements. Understanding human mobil-ity in an urban environment is crucial to design mobility pat-terns that meet social behaviors and scale to the requirementsof modern smart cities [20]. CrowdSenSim includes a numberof human mobility patterns designed for pedestrian mobilityin urban environments.

Costs of Sensing: The sensing activity impacts the energybudget of the participants’ mobile devices. CrowdSenSim isable to capture the energy directly spent for the sensingtasks as well as the energy spent for communications. IoTand mobile devices are equipped with several communicationtechnologies, including 3G/LTE, WiFi and Bluetooth. Batteryusage of the mobile devices differs with respect to the commu-nication technology, and can have associated costs (e.g., usershave a limited monthly plan) [21].

IV. THE CROWDSENSIM ARCHITECTURE

The architecture of CrowdSenSim follows the design spec-ifications illustrated in Section III, implementing independentmodules to characterize the urban environment, the usermobility, the communication and the crowdsensing inputs,which depend on the application and specific sensing paradigmutilized. Fig. 2 shows graphically the relations between themodules, and Table I lists description of symbols that areexplained in detail hereafter.

A. City Layout Module

The module in charge of defining the city layout allowsthe researcher to input into the simulator the city wheresimulations will be performed. Specifically, the layout of thecity is defined in terms of a set of coordinates C containing

TABLE ISYMBOLS LIST AND DESCRIPTION

SYMBOL DESCRIPTION

C Set of coordinates defining the city layoutTmove Amount of time each user moves in the citySmove Velocity of user movementca Coordinate where a user starts movingta Time when a user starts moving from cacnext Next coordinate of user movementtnext Time of arrival in the next coordinate cnextttravel Time necessary to move between two coordinatesE Energy spent for communication purposesPt x Power consumed for transmission over WiFi link

User MobilityList of Events

City Layout

SIMULATORCrowdSensing

Inputs

Results

Fig. 2. Main modules of CrowdSenSim

information on <latitude, longitude, altitude>. The set ofcoordinates compose the streets of the city where the userswill move during simulation runtime and can be obtained withonline tools like OpenStreetMaps or DigiPoint. In this versionof the simulator, we rely on Digipoint, which is a crowd-sourced application providing free access to street-level maps[22]. Fig. 3 shows the urban environments currently availablefor simulations, namely the city center of Luxembourg (seeFig. 3(a)), Trento (see Fig. 3(b)) and Madrid (see Fig. 3(c)).The center of Luxembourg city covers an area of 1.11 km2

with a population of 110 499 inhabitants as of the end of 2015and is the home of many national and international institutionalbuildings. The city center of Trento occupies an area of 1.18km2 and has a population of 117 317 inhabitants as of thebeginning of 2016 and is the capital of the homonym Province.The city center of Madrid covers approximately an area of 5.23km2 with a resident population of 149 718 residing inhabitants.

The city layout module allows the researcher to define thesize of the city and the level of detail of the urban environment.High resolution of the city layout, which corresponds tochoose a higher number of coordinates, increases the precisionof user movements at the cost of longer and more computation-ally expensive simulations. Viceversa, a coarse resolution ofthe city layout makes the simulations to run faster, but lowersthe accuracy of users movements and precision of the urbanenvironment. The latter component is important: having a highresolution of the urban environment permits to characterizeplaces, e.g., to identify among the others bars, restaurants,



IEEE ACCESS 4

(a) Luxembourg (b) Trento (c) Madrid

Fig. 3. Maps of cities obtained from DigiPoint

schools or hospitals.

B. User Mobility Module

The user mobility module defines the spatiotemporal prop-erties of user movements in the urban environment, whichcompose the so-called list of events (see Fig. 2). We definean event as “the arrival of an user in a given coordinate at agiven instant of time.”

The module defines the following steps to determine thespatiotemporal list of events:• Initialization: it characterizes the location and time of user

arrival.• Mobility: it characterizes the user movements after arrival.1) Initialization: This initial step is in charge of deter-

mining where and when each user starts moving in the city.Each user arrival is therefore characterized by a coordinateca and time ta. In the current version of the simulator, thelocation is randomly determined among the set of coordinatesC of the map. The design choice builds on the assumptionthat each of the coordinates has the same relevance, i.e.,it does not exist a difference between popularity of places.Future implementations will allow the researchers to choosebetween random and popularity-driven assignment of userlocation. The time of user arrival can be either randomizedor based on real-world traces, which are the results of a studyon pedestrian mobility and are public available on Crawdad(ostermalm_dense_run2) [23]. Fig. 4 shows the probabilitydensity function of the user arrival resulting from the studyof the traces. In practice, to obtain the results presentedlater in Section V-A2, the density computed in Fig. 4 wasadapted to an arrival time period between 8:00 AM - 1:40 PMinstead of 720 s and for 20 000 users. The probability densityfunction of user arrival is indeed determined by two globalsimulation inputs: the total number of users in the systemand the simulation period. In random user arrival modes, thedefault probability density function is uniform, i.e., duringthe simulation period each minute has the same probabilityto be chosen as arrival time for each user. The researcher caneasily modify the user arrival time by changing the probabilitydensity function. In the case study presented in Section VI, wewill present a modification of the probability density functionof user arrival suitable for the application of public streetlighting.

2) Mobility: In the default setting, each user moves overthe set of coordinates C for a predefined amount of timeTmove which is uniformly distributed between [10, 20] minutes

0

50 100

150

200

250

300

350

400

450

500

550

600

650

700

25

50

75

100

125

150

175

Time Interval (sec)

Den

sity

Fig. 4. User distribution of mobility trace “kth/walkers”

with an average speed Smove uniformly distributed between[1, 1.5] m/s. The default setting can be easily modified. Afterthe arrival in ca at time ta, the next move makes the user tojump in cnext at time tnext. The simulator chooses cnext to bephysically in proximity of ca, i.e., CrowdSenSim chooses acoordinate among C which is on the same street or squarewith distance below a maximum radius. No obstacles areconsidered between the move from one coordinate to anotherone. Once cnext has been determined, the simulator computestnext on the basis of the physical distance between ca andcnext and the speed of the user. The distance is computed byusing the Haversine formula [27] and, along with the speedof the movements, permits to determine the amount of time ittakes between the two points ttravel. Then, tnext is determinedas follows:

tnext = ta + ttravel, (1)

and the total amount of time the user is allowed to travel Tmoveis updated as follows:

Tmove = Tmove − ttravel. (2)

The user stops moving when Tmove ≤ 0. It is worth to highlightthat during each movement the speed of the movement Smovechanges. The new value is generated again uniformly dis-tributed between [1, 1.5] m/s to mimic the change of velocityduring walking.

In the current version, users move only once during thesimulation period, and it is not possible yet to define adirection of movement for each user. We plan to extend thesimulator to take into account this possibility in the futureextension of this study.



IEEE ACCESS 5

TABLE IISENSOR AND COMMUNICATION EQUIPMENT PARAMETERS USED FOR PERFORMANCE EVALUATION

SENSOR PARAMETER VALUE UNIT

Accelerometer Sample rate 50 HzSample size 12 BitsCurrent 35 µA

Temperature Sample rate 182 HzSample size 16 BitsCurrent 182 µA

Pressure Sample rate 157 HzSample size 16 BitsCurrent 423.9 µA

(a) Sensor Equipment [24], [25]

SYMBOL VALUE UNIT DESCRIPTION

ρid 3.68 W Power in idle modeρt x 0.37 W Transmission powerρr x 0.31 W Reception powerλg 1000 fps Rate of generation of packetsγxg 0.11 · 10−3 J Energy cost to elaborate a generated packet

(b) Communication Equipment [26]

C. Crowdsensing Inputs Module

This module defines the inputs specific to crowdsensinganalysis. CrowdSenSim relies on two types of inputs. The firstset does not depend on the sensing paradigm employed andcomprises all the parameters related to sensing and commu-nication operations. The second set includes parameters thatare specific to the participatory sensing paradigm. Unlike theopportunistic sensing paradigm which does not have particularinput parameters, in participatory systems it is necessary todefine the concept of task and how to assign tasks to users.

Sensing and Communication Parameters: In CrowdSenSim,data generation takes into account sensors commonly availablein current IoT and mobile devices. Table II presents the de-tailed information on sensors and communication parameters.Specifically, CrowdSenSim generates sensing readings fromthe FXOS8700CQ 3axis linear accelerometer from FreescaleSemiconductor [24] and the BMP280 from Bosch [25], whichis a digital pressure and temperature sensor. For a worstscenario analysis, in the default settings the sensors keepgenerating data according to their sampling frequency for theentire period of users movements.

For communication purposes, the current version of thesimulator employs only WiFi technology. Based on the sampleresolution of the sensors, data is first organized in packets of1 500 Bytes and delivered to the collector continuously duringusers movements. Each user transmits data to the closes WiFiAccess Point (AP). The APs are characterized by <latitude,longitude>, not necessarily from the set C. For the city ofLuxembourg, the precise location of WiFi APs was obtainedfrom an online tool1.

Parameters for Participatory Sensing Paradigm: Crowd-SenSim defines the following properties for tasks: location,time of deployment, duration and coverage. With the defaultsettings, all the parameters are randomly selected from theset of coordinates C, uniformly distributed within the sim-ulation period and as fraction of the simulation period forlocation, time of deployment and duration respectively. Thetask coverage defines the maximum radius where users canactively contribute to the task and is fixed for all the tasks.The researcher can also provide a file in input to the simulatordescribing the aforementioned properties.

1Online: https://www.hotcity.lu/en/laptop/www/About/Wi-Fi-coverage

D. Simulator and Results

During simulation CrowdSenSim computes runtime a num-ber of statistics, including energy consumption and amount ofdata generated and provides the researcher to a visualizationtool to display the results. For example, with the help ofGoogle Heatmap tool2, CrowdSenSim draws on the real mapsthe most populated tasks or most utilized WiFi APs. To illus-trate considering the former case as an example, CrowdSenSimcollects statistics about the number of users recruited for eachtask. At the end of the simulation period, it outputs thesestatistics along with the location of each task in terms oflatitude and longitude. The result obtained is then employedas input of the Google Heatmap tool (see Fig. 5).

The energy E spent for communication purposes is com-puted as follows. E is consumed during a transmission timeτtx and is defined as:

E =∫ τt x

0Ptx dt, (3)

where Ptx is the power consumed for transmissions of WiFipackets generated at rate λg [26]:

Ptx = ρid + ρtx · τtx + γxg · λg . (4)

V. PERFORMANCE EVALUATION

This section provides performance analysis of CrowdSen-Sim. First, the results obtained for participatory and op-portunistic sensing systems are illustrated, with a focus onparticipant recruitment for the former sensing paradigm andenergy consumption and amount of data collected for thelatter sensing paradigm. Second, technical evaluation of thesimulator is shown, with a focus on CPU, processing timeand memory utilization.

For performance evaluation, the simulations are carriedout using a Linux workstation equipped with Ubuntu 14.10.Furthermore, the machine supports an Intel ®Core TM i3 2.27GHz CPU and a system memory of 1916 MiB.

A. Analysis of Participatory and Opportunistic CrowdsensingScenarios

1) Participatory Sensing Scenario: In the participatorysensing scenario, we employ CrowdSenSim in the context of

2Online: https://developers.google.com/maps/documentation/javascript/examples/layer-heatmap



IEEE ACCESS 6

0

5

10

15

20

25

30

35

40

(a) SDRP Policy0

5

10

15

20

25

30

35

40

(b) DBRP Policy

Fig. 5. User recruitment for sensing tasks deployed in the city of Luxembourg

TABLE IIISIMULATION SETTINGS FOR ANALYSIS OF PARTICIPANT RECRUITMENT

POLICY

PARAMETER VALUE

Number of users [10 000]Overall evaluation period 8:00 AM - 2:00 PMTime of travel per user Uniformly distributed in [10, 20] minAverage user velocity Uniformly distributed in [1, 1.5] m/s

Timeslot duration 1 minuteTask duration 30 timeslotsNumber of tasks 25Dmax 30 m

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

0

5

10

15

20

25

30

35

40

45

Task ID

Num

bero

fUse

rs

SDRM DBRM

Fig. 6. Number of recruited users using SDRM and DBRM

participant recruitment and to implement a policy defininguser recruitment and task assignment [15]. Devising properrecruitment policy is important: on one hand, it allows theorganizer to minimize the expenditure, while on the otherhand, it helps to choose the users that will successfullycarry out the sensing task. For example, in the public safetycontext, it is essential to select users to maximize the trust-worthiness of collected data [28]–[30]. Such policy can beemployed using distance-based recruitment mode (DBRM) orsociability-driven recruitment mode (SDRM). In DBRM, thespatial distance between the users and the sensing task is thediscriminant factor defining user eligibility. Users far from the

sensing task i more than Dmax are never considered as potentialcontributors in that task. In SDRM, the user sociability, definedas amount of data users consume or the time they spend usingmobile social network applications [31], is the discriminantfactor for the recruitment.

Table III lists the details of the simulation set-up. Weemploy CrowdSenSim for demonstration purposes to visualizethe distribution of user recruitment and refer the reader forfurther details on the results to [15]. Fig. 5 compares thenumber of users recruited in SDRM and DBRM for all thedeployed 25 tasks in Luxembourg city center using the GoogleHeatmaps tool. Tasks with higher number of users recruitedare marked with a bigger radius and with brighter and moreintense colors. Fig. 6 shows that SDRM outperforms DBRMas the number of recruited users is higher for all the deployedtasks. Moreover, for task with ID equal to 8, the SDRM isable to recruit users where the DBRM fails.

2) Opportunistic Sensing Scenario: In the opportunisticsensing scenario, users contribute continuously data even ifthey do not receive a specific task. In this context, Crowd-SenSim is employed for evaluation of data generation with afixed the number of participants set to 20 000. The objective ofthe experiment is to assess during the simulation period from8:00 AM to 2:00 PM the energy consumption attributed tosensing and reporting operations and the amount of generateddata. The analysis is carried under the two different user arrivalpatterns. Users move according to the predefined settingsillustrated in Section IV-B. In the first user arrival pattern,the starting time of the walk is uniformly distributed between8:00 AM and 1:40 PM to allow users starting moving towardsthe end of the period to correctly end their journey at 2:00 PM.The second arrival pattern is based on the data set with tracesof pedestrian mobility (ostermalm_dense_run2) [23].

Energy Cost for Sensing and Reporting: Fig. 7 presentsthe distribution of users and their energy spent for sensingwith the uniform and traces-based user arrival patterns. Fordemonstration purposes, we show the results obtained forthe sole city of Luxembourg. As expected, the user arrivalpattern does not influence the energy consumption, which onlydepends on the amount of time the users generate data. As theusers contribute data for time periods as low as 10 minutes up



IEEE ACCESS 7

0-50

50-1

0010

0-15

015

0-20

020

0-25

025

0-30

030

0-35

035

0-40

040

0-45

045

0-50

050

0-55

055

0-60

060

0-65

065

0-70

070

0-75

075

0-80

080

0-85

085

0-90

0

0500

100015002000250030003500

Current drain (µAh)

Num

.ofP

artic

ipan

ts Uniform Traces

(a) Sensing Cost

0-5

5-10

10-1

515

-20

20-2

525

-30

30-3

535

-40

40-4

545

-50

50-5

555

-60

60-6

565

-70

70-7

575

-80

80-8

5

0500

100015002000250030003500

Energy (J)

Num

.ofP

artic

ipan

ts Uniform Traces

(b) Communication Cost

Fig. 7. Energy spent for sensing and communication

to time periods of a maximum of 20 minutes, the profiles ofFig. 7(b) and Fig. 7(a) follow a normal distribution. Currentdrain of sensing operations is on average 373.41 µAh and368.80 µAh for uniform and traces-based arrival patterns. Inthe worst case, few users experience a cost that is nearly morethan double with respect to the average. Comparing to thebattery capacity available in modern smartphones, which isin the order of 2 000 mAh, it is possible to conclude that theenergy cost for sensing is negligible with respect to the energyspent for communications (see Fig. 7(b)).

Amount of Data Collected: The amount of informationreported by users devices is unveiled in the following ex-periments, which evaluate the amount of data generated persingle sensor for the two different user arrival patterns and thedistribution of the data collected.

Fig. 8 shows the total amount of data collected along withthe simulation period for the two user arrival patterns. Asexpected, the amount of data is proportional to the samplingfrequencies of the three considered sensors. Recalling thateach user contributes only during a short period of time (from10 to 20 minutes), the amount of collected information isconsiderable. For example, 20 000 users arriving according tothe uniform arrival pattern would generate 2.62 GiB, 12.71GiB and 10.96 GiB for the accelerometer, temperature andpressure sensors respectively. Fig. 8(a) shows the results forthe uniformly distributed arrival pattern. As expected, theamount of contribution remains constant after the initial setup as the number of users arriving in a given time windowis constant along the simulation period. Fig. 8(b) illustratesthe results for the user arrival pattern based on the data set.Unlike the previous case, the shape of the curve follows theprobability density function of the traces as in Fig. 4.

Having the knowledge on the amount of data the userscan contribute is important, but for more precise evaluationit is also fundamental to determine where and when thesesamples are generated. CrowdSenSim provides the researchersthe capability to graphically visualize the data generationprocess. With a number of users set to 20 000, Fig. 9 showsthe geographical distribution of the amount of collected dataat the end of the simulation period for Luxembourg, Trentoand Madrid. To better analyze the data generation process, wedefine a new performance metric, called Sample Distribution

Coefficient (SDC), which measures the amount of generatedsamples per meter and is defined as follows:

SDC =Nt

∆, (5)

where ∆ is the average distance between samples and Nt isthe number of samples generated during the time period t. Theparameter ∆ is defined as follows:

∆ =

∑ni, ji≥ j

d(i, j)

n(n − 1)2

. (6)

The term d(i, j) is the distance (in meters) between the locationwhere the samples i and j were generated and the denominatoraccounts for the number of pairs of samples. SDC can becomputed at any temporal and spatial resolution. The timegranularity can be fine or coarse, e.g., minute, hour or daywhereas the spatial granularity can be at block-, district- oreven city-level. For example, SCD can be employed to analyzethe per-hour data generation process in a downtown district vssuburban district.

Fig. 10 shows the distribution of SDC for Luxembourg,Trento and Madrid for the entire simulation period. In thisexperiment, the users are located with the uniform arrivalpattern. It is interesting to notice that the lowest values ofSDC occur for the initial and final time intervals (8:00 AM- 9:00 AM and 1:00 PM - 2:00 PM). During the initial andfinal time intervals the number of participants is lower thanin the other intervals as the simulator locates the users witha uniform distribution between 8:00 AM and 1:40 PM andthey move for at maximum 20 minutes. Having set the samenumber of users for the experiment, the relation between theSDC coefficient and the size of the area considered is inverselyproportional. The city center of Luxembourg is smaller thanTrento and Madrid. As a result, the obtained SDC value forLuxembourg is higher.

B. Performance of the Simulator

This section provides a technical evaluation of the simulatorperformance. The metrics evaluated concern processing time,CPU and memory utilization.



IEEE ACCESS 8

8:00

8:20

8:40

9:00

9:20

9:40

10:0

010

:20

10:4

011

:00

11:2

011

:40

12:0

012

:20

12:4

013

:00

13:2

013

:40

14:0

0

0

0.2

0.4

0.6

0.8

Time

Dat

a(G

iB)

Accelerometer Temperature Pressure

(a) Arrival pattern with uniform distribution

8:00

8:20

8:40

9:00

9:20

9:40

10:0

010

:20

10:4

011

:00

11:2

011

:40

12:0

012

:20

12:4

013

:00

13:2

013

:40

14:0

0

0

0.5

1

1.5

2

Time

Dat

a(G

iB)

Accelerometer Temperature Pressure

(b) Arrival pattern based on traces

Fig. 8. Amount of data generated

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

(a) Luxembourg0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

(b) Trento0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

(c) Madrid

Fig. 9. Normalized distribution of amount of collected data for the different cities over the time period 8:00 AM - 2:00 PM

8:00-9

:00

9:00-1

0:00

10:00

-11:00

11:00

-12:00

12:00

-13:00

13:00

-14:00

0

50

100

150

200

Time interval

SDC

(sam

ple/

m)

(a) Luxembourg

8:00-9

:00

9:00-1

0:00

10:00

-11:00

11:00

-12:00

12:00

-13:00

13:00

-14:00

0

50

100

150

200

Time interval

SDC

(sam

ple/

m)

(b) Trento

8:00-9

:00

9:00-1

0:00

10:00

-11:00

11:00

-12:00

12:00

-13:00

13:00

-14:00

0

50

100

150

200

Time interval

SDC

(sam

ple/

m)

(c) Madrid

Fig. 10. Sample Distribution Coefficient for the different cities over the time period 8:00 AM - 2:00 PM

Fig. 12 shows the profile of the CPU utilization expressedin percentage obtained with the dstat tool3. The experimentanalyzes the performance in a scenario with a huge numberof users, 100 000, in the city of Luxembourg. The statisticsobtained have been filtered to spot the profile of the processrunning the simulation. The resulting graph shows that theCPU utilization can occupy as much as 25% of the availableresources and this happens at the beginning where most of thecomputation occurs to process the events.

The next set of experiments aims at assessing the perfor-mance of processing time and memory occupancy. Unlikethe previous result, these experiments are carried out de-ploying CrowdSenSim in a Virtual Machine (VM) runningUbuntu 14.10 with two different profile settings, namely

3Available on: http://dag.wiee.rs/home-made/dstat/

1024 MiB and 2048 MiB of memory. The setting allowsus to profile the performance of the simulator perceived bythe end users. The VM is equipped with GNOME Sys-tem Monitor which permits to verify the system perfor-mance. Fig. 11 shows an example for a simulation with20 000 participants in opportunistic sensing scenario. Fig. 13shows the results obtained. Both experiments were performedfor the city of Luxembourg, with both VMs configurationsand with an increasing number of participants from theset {1 000, 5 000, 10 000, 20 000, 50 000, 70 000, 100 000}. Themaximum number of users was selected consistently with thepopulation of the city. Fig. 13(a) analyzes the processing time,which remains almost constant for a number of participantslower than 10 000 and then it increases exponentially for boththe configuration settings. Fig. 13(b) analyzes the memory



IEEE ACCESS 9

Fig. 11. GNOME System Monitor

0 25 50 75 100 125 150 175 200 2250

5

10

15

20

25

Simulation Period (s)

CPU

%

Fig. 12. CPU utilization for a simulation run with 100 000 users

consumption with a focus on the Resident Set Size (RSS),which defines the amount of memory the process occupiesin the RAM. For both configurations of the VM, the RSSremains almost identical for a number of participants lowerthan 20 000, then the process tends to occupy as much aspossible all the available resources.

VI. CASE STUDY: SMART LIGHTING

CrowdSenSim is a candidate tool for analysis of smartcity services. This section presents a case study where thesimulator is employed to assess the performance of publicstreet lighting. However, the capabilities of the simulatorare not restrained to this particular application scenario. Weare currently working to extend the simulator capabilities toinclude vehicles as contributors to the data collection processand to analyze other important and challenging issues ofmodern cities, e.g., waste management. Waste managementinvolves the whole process of monitoring waste locations,truck routes, collection phases and waste disposal.

A. The problem of Smart Lighting in Modern Smart Cities

Public lighting is a traditional city service provided bylampposts widely distributed in streets and roads. Lightingcauses nearly 19% of worldwide use of electrical energyand entails a 6% of global emissions of greenhouse gases.A decrease of 40% of energy spent for lighting purposesis equivalent to eliminate half of the emissions from theproduction of electricity and heat generation of the US [32].Specifically, public street lightning, which is an essential com-munity service, impacts for around 40% on the cities’ energybudget. Consequently, in preparation of the EU commitments,optimizing the lighting service is a primary objective for themunicipalities [33].

The street lighting solutions currently implemented in citiesare not energy efficient. Typically, every lamp operates atfull intensity 12 hours a day on average: 8 hours duringsummer and 14 hours during winter period [33]. As a result,the costs the municipalities sustain are high [32]. A numberof different types of lamps are applicable for public streetlighting, including High Pressure Sodium (HPS), Metal-halide(MH) lamps, Compact Fluorescent lamps (CFL) and Light-emitting diode (LED). LEDs have an average lifetime 4 timeslonger than HPS lamps and 10 times longer if compared toMH lamps. Installing LEDs is effective to reduce hardware,installation and maintenance costs. Low wattage providessignificant energy savings and allows increasing the lampefficiency [34], [35]. The HPS lamps do not support dimmingand only LEDs can be employed to perform dimming properly.The use of LEDs is gradually gaining popularity due to itsphoto metric characteristics, such as low weighted energyconsumption (kW/1000hrs), high luminous efficacy (lm / W),high mechanical strength, long lifespan and reduction of lightpollution. LED lamps can dim the light intensity by more than50% modifying therefore the output level of light accordingto the circumstances. For example, when traffic is low or inrarely visited areas of the city, like the parks at night. The



IEEE ACCESS 10

1 000 10 000 100 0000

10

20

30

40

Number of Users

Proc

essi

ngTi

me

(s) 1024 MB

2048 MB

(a)

1 000 10 000 100 0000

500

1 000

1 500

2 000

Number of Users

RSS

(MB

)

1024 MB2048 MB

(b)

Fig. 13. Analysis of a) processing time and b) memory with increasing number of users

R

Fig. 14. Coverage radius R. The presence of sensors at the lampposts enablesrecognition of citizens within a radius of R.

Fig. 15. Position of lampposts in Luxembourg city center

city of Brittany in France, dims street lights by 60% between11:00 PM and 5:00 AM to decrease waste energy [33].

We devise a smart lighting method for smart cities whichdims the light of lampposts in proportion to the number ofusers in the vicinity. To detect the presence of users nearbythe lampposts a presence sensor like the SE-10 PIR motionsensor is assumed to be installed on site [36]. With presencesensors, every lamppost is able to recognize the presence ofcitizens within a certain radius R like illustrated in Fig. 14.Similarly to the solution adopted in Brittany, i.e., the minimumlight intensity level is 60% if no users are within the coverageradius R and increases or decreases proportionally on the basisof the passage of the users. In more details, if the number ofusers is increasing, then the light intensity increases or remainsat 100%, while if the number of users reduces from previousstatus, then the light intensity diminishes until it reaches the

PM AM

9-10 10-11 11-12 12-1 1-2 2-3 3-4 4-5 5-6 6-70

0.1

0.2

0.3

Time Interval (hour)

PDF

Fig. 16. Probability density function of user mobility during the evaluationperiod. Probability of a user to change their location is higher in early morningor late evening hours.

minimum level.

B. Evaluating Smart Lighting Solutions with CrowdSenSim

To evaluate the proposed smart lighting solution withCrowdSenSim, a set of 537 lampposts has been deployedaccording to their physical location in the streets and squaresof Luxembourg City. Fig. 15 details the position of eachlamppost given in terms of coordinates <latitude, longitude,altitude>. We compare two cases. In the proposed smart light-ing solution each lamp is equipped with LED technology andat full light intensity consumes 82.7 kW/1000hrs. In currentimplementation, each lamp is equipped with HPS technologyconsuming 172.7 kW/1000hrs at full light intensity.

The number of users moving in the city is set to 5 000. Eachof them walks for a period of time that is uniformly distributedin [10, 20] minutes with an average speed uniformly distributedbetween [1, 1.5] m/s. The users begin walking according toa specific arrival pattern. During the evaluation period, setbetween 9:00 PM and 7:00 AM, each user has a probabilityto start traveling that is defined by the probability densityfunction (PDF) illustrated in Fig. 16. In more details, during9:00 PM and 10:00 PM nearly one third of the total numberof users starts walking and at 7:00 AM all 5 000 users endtraveling.

Fig. 17 shows the results of the lamppost activity obtainedthrough CrowdSenSim. On average, the smart lighting solutionwith LED technology and light dimming saves nearly 68%of energy consumption with respect to the current adopted



IEEE ACCESS 11

50

55

60

65

70

75

80

85

90

(a) Lampposts activity with LEDs

50

75

100

125

150

175

200

(b) Lampposts activity with current technologymethod

Fig. 17. Heatmap of lampposts activity (values in kWh)

solution. Indeed, the set of lampposts consumes on average298.5 kWh per day with dimming and a fix amount of energyof 927.4 kWh per day with current implementation.

VII. CONCLUSION

In this paper we presented CrowdSenSim, a simulationplatform for MCS systems. CrowdSenSim is tailored to assesssensing activities in large-scale realistic urban environmentsand is designed to output results on participant recruitment,data generation and the cost sustained for sensing and report-ing from the users point of view. We also demonstrated thesuitability of the simulator for analysis of smart city serviceswith a case study on public street lighting. CrowdSenSim isdistributed as public available software.4

For future work, we plan to validate simulation resultsCrowdSenSim generates with experimental data obtained fromexisting crowdsensing platforms. Future development direc-tions are twofold. First, we plan to implement a more so-phisticated and accurate communication model to analyze inmore details the networking aspects of MCS systems. Second,we plan to develop a function to allow researchers to definedirections of user movements on individual basis. Futureresearch directions will exploit CrowdSenSim to investigateother important city services such as smart waste managementand extend the simulator to operate in vehicular environment,where vehicles contribute to the process of data generation inaddition to mobile devices. The current trend sees automotivecompanies to increase on-board equipment of vehicles with

4Available on: http://crowdsensim.gforge.uni.lu/

storage, computing capabilities and a growing set of sensors.Data collected by these sensors is not only beneficial for theoperation of the vehicles and monitoring of their status, butis projected to become a precious source of information formunicipalities as well.

REFERENCES

[1] N. B. Grimm, S. H. Faeth, N. E. Golubiewski, C. L. Redman, J. Wu,X. Bai, and J. M. Briggs, “Global change and the ecology of cities,”in Science, vol. 319, no. 5864, 2008, pp. 756–760. [Online]. Available:http://www.sciencemag.org/content/319/5864/756.abstract

[2] H. B. Dulal and S. Akbar, “Greenhouse gas emission reduction optionsfor cities: Finding the “coincidence of agendas” between local prioritiesand climate change mitigation objectives,” Habitat International, vol. 38,pp. 100 – 105, 2013.

[3] A. Caragliu, C. Del Bo, and P. Nijkamp, “Smart cities in europe,”Journal of urban technology, vol. 18, no. 2, pp. 65–82, 2011.

[4] A. Zanella, N. Bui, A. Castellani, L. Vangelista, and M. Zorzi, “Internetof Things for smart cities,” IEEE Internet of Things Journal, vol. 1,no. 1, pp. 22–32, Feb 2014.

[5] C. Perera, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Sensingas a service model for smart cities supported by Internet of Things,”Transactions on Emerging Telecommunications Technologies, vol. 25,no. 1, pp. 81–93, 2014.

[6] A. D. Cartier, D. H. Lee, B. Kantarci, and L. Foschini, “IoT-Big DataSoftware Ecosystems for Smart Cities Sensing: Challenges, Open Issues,and Emerging Solutions,” in 4th International Workshop on Cloud forIoT (CLIoT), 2016, p. (accepted).

[7] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, andM. Ayyash, “Internet of Things: A survey on enabling technologies,protocols, and applications,” IEEE Communications Surveys Tutorials,vol. 17, no. 4, pp. 2347–2376, Fourth quarter 2015.

[8] G. Cardone, A. Cirri, A. Corradi, L. Foschini, R. Ianniello, and R. Mon-tanari, “Crowdsensing in urban areas for city-scale mass gathering man-agement: Geofencing and activity recognition,” IEEE Sensors Journal,vol. 14, no. 12, pp. 4185–4195, Dec 2014.

[9] J. G. P. Rodrigues, A. Aguiar, and J. Barros, “Sensemycity: Crowdsourc-ing an urban sensor,” CoRR, vol. abs/1412.2070, 2014.

[10] A. Antonic, V. Bilas, M. Marjanovic, M. Matijaševic, D. Oletic,M. Pavelic, I. P. Žarko, K. Pripužic, and L. Skorin-Kapov, “Urbancrowd sensing demonstrator: Sense the Zagreb air,” in 22nd InternationalConference on Software, Telecommunications and Computer Networks(SoftCOM), Sept 2014, pp. 423–424.

[11] R. Ganti, F. Ye, and H. Lei, “Mobile crowdsensing: current state andfuture challenges,” IEEE Communications Magazine, vol. 49, no. 11,pp. 32–39, November 2011.

[12] W. Khan, Y. Xiang, M. Aalsalem, and Q. Arshad, “Mobile phone sensingsystems: A survey,” IEEE Communications Surveys Tutorials, vol. 15,no. 1, pp. 402–427, First 2013.

[13] P. P. Jayaraman, J. B. Gomes, H. L. Nguyen, Z. S. Abdallah, S. Krish-naswamy, and A. Zaslavsky, “Scalable energy-efficient distributed dataanalytics for crowdsensing applications in mobile environments,” IEEETransactions on Computational Social Systems, vol. 2, no. 3, pp. 109–123, Sept 2015.

[14] Google Inc., “Google science journal,” 2016. [Online]. Available:https://makingscience.withgoogle.com/science-journal/

[15] C. Fiandrino, B. Kantarci, F. Anjomshoa, D. Kliazovich, P. Bouvry,and J. Matthews, “Sociability-driven user recruitment in mobile crowd-sensing internet of things platforms,” in IEEE Global CommunicationsConference (GLOBECOM), Dec 2016, pp. 1–6.

[16] K. Farkas and I. Lendák, “Evaluation of simulation engines for crowd-sensing activities,” in 3rd International Conference & Workshop Mecha-tronics in Practice and Education, ser. MECHEDU, 2015, pp. 126–131.

[17] C. Tanas and J. Herrera-Joancomartí, “Crowdsensing simulation usingns-3,” Citizen in Sensor Networks: Second International Workshop,CitiSens 2013, pp. 47–58, 2014, Springer International Publishing.

[18] K. Farkas and I. Lendák, “Simulation environment for investigatingcrowd-sensing based urban parking,” in International Conference onModels and Technologies for Intelligent Transportation Systems (MT-ITS), June 2015, pp. 320–327.

[19] K. Mehdi, M. Lounis, A. Bounceur, and T. Kechadi, “Cupcarbon:A multi-agent and discrete event wireless sensor network design andsimulation tool,” in 7th International ICST Conference on SimulationTools and Techniques, ser. SIMUTools ’14, 2014, pp. 126–131.



IEEE ACCESS 12

[20] K. K. Jahromi, M. Zignani, S. Gaito, and G. P. Rossi, “Simulating humanmobility patterns in urban areas,” Simulation Modelling Practice andTheory, vol. 62, pp. 137 – 156, 2016.

[21] C. Fiandrino, D. Kliazovich, P. Bouvry, and A. Y. Zomaya, “Network-assisted offloading for mobile cloud applications,” in IEEE InternationalConference on Communications (ICC), June 2015, pp. 5833–5838.

[22] Zonum SOlutions, “Digipoint 3,” 2014. [Online]. Available: http://www.zonums.com/gmaps/digipoint.php

[23] S. T. Kouyoumdjieva, Ólafur Ragnar Helgason, and G. Karlsson,“CRAWDAD dataset kth/walkers (v. 2014-05-05),” Downloaded fromhttp://crawdad.org/kth/walkers/20140505, May 2014.

[24] “FXOS8700CQ: Digital Sensor - 3D Accelerometer + 3DMagnetometer,” http://cache.nxp.com/files/sensors/doc/data_sheet/FXOS8700CQ.pdf?pspll=1, 2015.

[25] “BMP280, Barometric Pressure Sensors,” https://www.bosch-sensortec.com/bst/products/all_products/bmp280, 2015.

[26] A. Garcia-Saavedra, P. Serrano, A. Banchs, and G. Bianchi, “Energyconsumption anatomy of 802.11 devices and its implication on modelingand design,” in 8th ACM International Conference on Emerging Net-working Experiments and Technologies, ser. CoNEXT, 2012, pp. 169–180.

[27] C. Robusto, “The cosine-haversine formula,” The American Mathemat-ical Monthly, vol. 64, no. 1, pp. 38–40, 1957.

[28] B. Kantarci and H. T. Mouftah, “Trustworthy sensing for public safetyin cloud-centric Internet of Things,” IEEE Internet of Things Journal,vol. 1, no. 4, pp. 360–368, Aug 2014.

[29] M. Pouryazdan, C. Fiandrino, B. Kantarci, D. Kliazovich, T. Soyata, andP. Bouvry, “Game-theoretic recruitment of sensing service providers fortrustworthy cloud-centric internet-of-things (iot) applications,” in IEEEGlobecom Workshops (GC Wkshps), Dec 2016, pp. 1–6.

[30] M. Pouryazdan, B. Kantarci, T. Soyata, and H. Song, “Anchor-assistedand vote-based trustworthiness assurance in smart city crowdsensing,”IEEE Access, vol. 4, pp. 529–541, 2016.

[31] F. Anjomshoa, M. Catalfamo, D. Hecker, N. Helgeland, A. Rasch,B. Kantarci, M. Erol-Kantarci, and S. Schuckers, “Mobile behaviometricframework for sociability assessment and identification of smartphoneusers,” in 2016 IEEE Symposium on Computers and Communication(ISCC), June 2016, pp. 1084–1089.

[32] M. Castro, A. J. Jara, and A. F. G. Skarmeta, “Smart lighting solutionsfor smart cities,” in 27th International Conference on Advanced Infor-mation Networking and Applications Workshops (WAINA), March 2013,pp. 1374–1379.

[33] “The Business Case for Smart Street Lights,”http://www.silverspringnet.com/wp-content/uploads/SilverSpring-Whitepaper-Smart-Street-Light-Bizcase.pdf, 2013,Accessed July 26, 2016.

[34] “Philips ClassicStreet,” http://download.p4c.philips.com/lfb/f/fp-912300023301/fp-912300023301\_pgl\_en\_aa\_001.pdf, 2016,Accessed July 26, 2016.

[35] “LucaloxTM Standard,” http://www.gelighting.com/LightingWeb/ru/images\\/HPS\_Lucalox\_Lamps\_Data\_sheet\_EN.pdf, 2016, AccessedJuly 26, 2016.

[36] F. Leccese, “Remote-control system of high efficiency and intelligentstreet lighting using a ZigBee network of devices and sensors,” IEEETransactions on Power Delivery, vol. 28, no. 1, pp. 21–28, Jan 2013.

Claudio Fiandrino (S’14) is a postdoctoral re-searcher at IMDEA Networks Institute, Madrid,Spain. Claudio obtained his Ph.D. degree at theUniversity of Luxembourg working on the ECO-CLOUD project focusing on energy efficient com-munications in cloud, mobile cloud and fog comput-ing. He received the Bachelor Degree in IngegneriaTelematica in 2010 and the Master Degree in Com-puter and Communication Networks Engineering in2012 both from Politecnico di Torino. Claudio’swork on indoor localization over fog computing

platforms received the Best Paper Award in IEEE CloudNet 2016. Claudiowas a Visiting Ph.D. Student for three months at Clarkson University, NY,USA. He served as Publication and Web Chair at the IEEE InternationalConference on Cloud Networking (CloudNet 2014) and as TPC memberin IEEE CloudNet 2014, IEEE ICC 2016 - SAC Cloud Communicationsand Networking and IEEE EmergiTech 2016. His primary research interestsinclude mobile cloud/fog computing, mobile crowdsensing and data centercommunication systems.

Andrea Capponi (S’16) is a Ph.D. student at theUniversity of Luxembourg. He received the BachelorDegree in Telecommunication Engineering and theMaster Degree in Telecommunication Engineeringboth from the University of Pisa, Italy. His primaryresearch interests are Mobile Crowdsensing, Internetof Things and Smart Cities.

Giuseppe Cacciatore is currently a MS student inTelecommunications Engineering at the Universityof Trento. He performed his master thesis as anintern at the University of Luxembourg.

Dzmitry Kliazovich (M’03-SM’12) is a Head ofInnovation at ExaMotive. He was a Senior Sci-entist at the Faculty of Science, Technology, andCommunication of the University of Luxembourg.Dr. Kliazovich holds an award-winning Ph.D. inInformation and Telecommunication Technologiesfrom the University of Trento (Italy). Dr. Kliazovichis a holder of several scientific awards from theIEEE Communications Society and European Re-search Consortium for Informatics and Mathematics(ERCIM). His works on cloud computing, energy-

efficiency, indoor localization, and mobile networks received IEEE/ACMBest Paper Awards. He coordinated organization and chaired a number ofhighly ranked international conferences and symposia, including the IEEEInternational Conference on Cloud Networking (CloudNet 2014). Dr. Klia-zovich is the author of more than 100 research papers. He is the AssociateEditor of the IEEE Communications Surveys and Tutorials and of the IEEETransactions of Cloud Computing journals. He is a Vice Chair of the IEEEComSoc Technical Committee on Communications Systems Integration andModeling. Dr. Kliazovich is a coordinator and principal investigator of theEnergy-Efficient Cloud Computing and Communications initiative funded bythe National Research Fund of Luxembourg. His main research activities arein the field of intelligent transportation systems, telecommunications, cloudcomputing, and Internet of Things (IoT).



IEEE ACCESS 13

Ulrich Sorger Ulrich Sorger is a professor at theComputer Science and Communication research unitof the Faculty of Science, Technology and Com-munication at the University of Luxembourg. Hisresearch interest include optimization, coding andinformation theory. Prof. Sorger holds a Habilitationin Communication Theory and a Ph.D. in ElectricalEngineering from the Technical University of Darm-stadt.

Pascal Bouvry is a professor in the ComputerScience and Communication research unit of theFaculty of Science, Technology and Communicationat the University of Luxembourg and a facultymember at the Luxembourg Interdisciplinary Cen-ter of Security, Reliability, and Trust. His researchinterests include cloud & parallel computing, opti-mization, security and reliability. Prof. Bouvry hasa Ph.D. in computer science from the University ofGrenoble (INPG), France. He is on the IEEE CloudComputing and Elsevier Swarm and Evolutionary

Computation editorial boards. He is also acting as communication vice-chairof the IEEE STC on Sustainable Computing and co-founder of the IEEE TCon Cybernetics for Cyber-Physical Systems. A full biography is available onpage http://pascal.bouvry.org. Contact him at [email protected].

Burak Kantarci (S’05–M’09–SM’12) is an Assis-tant Professor with the School of Electrical Engi-neering and Computer Science at the University ofOttawa (Ontario, Canada). From 2014 to 2016, hewas an Assistant Professor at the ECE Departmentat Clarkson University, where he currently holdsa courtesy appointment as an assistant professor.Dr. Kantarci received the M.Sc. and Ph.D. degreesin computer engineering from Istanbul TechnicalUniversity, in 2005 and 2009, respectively. He re-ceived the Siemens Excellence Award in 2005 for

his studies in optical burst switching. During his Ph.D. study, he studiedas a Visiting Scholar with the University of Ottawa, where he completedthe major content of his thesis. He has co-authored over 100 papers inestablished journals and conferences, and contributed to 11 book chapters.He is the Co-Editor of the book entitled Communication Infrastructures forCloud Computing. He has served as the Technical Program Co-Chair of seveninternational conferences/symposia/workshops. He is an Editor of the IEEECommunications Surveys and Tutorials. He also serves as the Secretary of theIEEE ComSoc Communication Systems Integration and Modeling TechnicalCommittee. He is a member of the ACM and a senior member of the IEEE.

Fabrizio Granelli (M’01–SM’05) is Associate Pro-fessor and Delegate for Education at the Dept. of In-formation Engineering and Computer Science (DISI)of the University of Trento (Italy) and IEEE ComSocDirector for Online Content. He received the «Lau-rea» (M.Sc.) degree in Electronic Engineering fromthe University of Genoa, Italy, in 1997, with a thesison video coding, awarded with the TELECOM Italyprize, and the Ph.D. in Telecommunications from thesame university, in 2001. Since 2000 he is carryingon his research and didactical activities (currently

Associate Professor in Telecommunications) at the Dept. of Information Engi-neering and Computer Science – University of Trento (Italy) and coordinatorof the Networking Laboratory. Between 2004 and 2015 for a total of sixmonths, he was visiting professor at the State University of Campinas (Brasil).In 2016, he was visiting professor at the University of Tokyo (Japan). Heis Associate Editor-in-Chief of IEEE Communications Surveys and Tutorialsjournal and he was IEEE ComSoc Distinguished Lecturer for the period 2012-15 (2 terms). He is author or co-author of more than 170 papers published ininternational journals, books and conferences, focused on networking.

Stefano Giordano (SM’10) received the master’sdegree in electronics engineering and the Ph.D. de-gree in information engineering from the Universitàdi Pisa, Pisa, Italy, in 1990 and 1994, respectively.He is currently an Associate Professor with theDepartment of Information Engineering, Universitàdi Pisa, where he is responsible for the telecommu-nication networks laboratories. His research inter-ests are telecommunication networks analysis anddesign, simulation of communication networks, andmultimedia communications. He is a member of the

Editorial Board of the IEEE COMMUNICATION SURVEYS AND TUTORIALS.He was the Chair of the Communication Systems Integration and ModelingTechnical Committee. He is an Associate Editor of the International Journalon Communication Systems and the Journal of Communication Softwareand Systems technically cosponsored by the IEEE Communication Society.He was one of the referees of the European Union, the National ScienceFoundation, and the Italian Ministry of Economic Development.

ieee access 1 crowdsensim: a simulation platform for

Documents